Geology Reference
In-Depth Information
85 GHz channel.
Andersen
[2000] conducted a study on
the sensitivity of ice concentration to cloud liquid water
and integrated water vapor. Sensitivities of the estimated
concentration to atmospheric influences are analyzed in
Andersen et al
. [2006]. Successful implementation of the
high‐frequency algorithms requires not only reliable
modules to correct for atmospheric influences but also
accurate input of atmospheric parameters, which are usu-
ally obtained from weather models. Atmospheric effects
are more pronounced over regions of OW and low ice
concentrations. For that reason, the retrieved concentra-
tions are almost always less accurate at the ice edge and in
marginal ice zones. This can be adequately improved
using radiative transfer models.
The sensitivity of ice concentration to perturbations
in emissivity or brightness temperature (i.e., the devia-
tion from their typical values) is usually examined
by incrementing the brightness temperature by a step
equivalent to the upper limit of the sensor's noise (usu-
ally 1 K) and calculating the ice concentration from a
given algorithm.
Markus and Cavalieri
[2000] used this
approach to determine the sensitivity of the NT2 algo-
rithm to variations in brightness temperature from
the three channels used in the algorithm (19, 37, and
85 GHZ from SSM/I). They conducted the calculations
using three different ice concentration values corre-
sponding to low, medium, and high concentration of
32%, 51%, and 98%, respectively. Except for the 19 GHz
horizontal observations (to which the ice concentration
has no sensitivity), the authors found that a change in
brightness temperature by 1 K from any frequency or
polarization channel led to a change in the output total
ice concentration between 3.2% and 6.5%. This was true
for the three ranges of ice concentration.
The same methodology was used to evaluate the sensi-
tivity of the calculated ice concentration from ECICE to
the passive microwave brightness temperature. Results
from using observations from
T
b
,85
h
and
T
b
,85
v
to determine
total ice concentration are presented in Figure 10.19.
Probability distributions of the input parameters for
ice (regardless of the type) and water were input to the
algorithm.
T
b
,85
h
was made to vary between 180 and 220 K
with a step of 2 K, while
T
b
,85
v
was varied between 230 and
250 K with a step of 1 K. The figure shows that total ice
concentration increases linearly with
T
b
,85
h
but non line-
arly with
T
b
,85
v
. For any given value of
T
b
,85
v
, the ice con-
centration increases by an average of 1.5% per one degree
increase in
T
b
,85
h
. The nonlinear variation of the calcu-
lated total concentration with
T
b
,85
v
is demonstrated fur-
ther in Figure 10.19b. At low values of
T
b
,85
h
, typical of
the radiometrically cold OW, ice concentration generally
increases at a rate of 0.67% per one degree increase in
T
b
,85
v
. At the higher end of
T
b
,85
h
(218 and 220 K in the
figure), which is typical of values from surface layering
and glaze [
Comiso et al
., 1997], an increase in concentra-
tion with
T
b
,85
v
is observed up to approximately 245 K,
followed by a decrease of 5% concentration per one
degree increase in
T
b
,85
v
.
10.2.2.3. Validation of Ice Concentration Algorithms
In order to evaluate the ice concentration accuracy
from any algorithm, a set of truth (validation) data must
be identified. The most common sources include opera-
tional ice charts, shipboard or field visual, airborne
(a)
(b)
90
80
70
60
T
b,
85
h
=180°K
T
b,
85
h
=182°K
T
b,
85
h
=218°K
T
b,
85
h
=220°K
50
40
30
20
10
0
225
230
235
240
T
b
,85
v
(°K)
245
250
255
Figure 10.19
Sensitivity of calculated total ice concentration from ECICE using two observations simulating
SSM/I,
T
b
,85
h
and
T
b
,85
v
, to variations in brightness temperature: (a) a 3D perspective and (b) cross sections of
2D display.
